Analysis of Genetic Diversity and the Determination of Relationships Among Western Mediterranean Horse Breeds Using Microsatellite Markers D

J. Anim. Breed. Genet. ISSN 0931-2668

ORIGINAL ARTICLE Analysis of genetic diversity and the determination of relationships among western Mediterranean horse breeds using microsatellite markers D. Marletta1, I. Tupac-Yupanqui2, S. Bordonaro1, D. Garcı´a2, A. M. Guastella1, A. Criscione3, J. Can˜ o´ n2 & S. Dunner2

1 DACPA, Sez. di Scienze delle Produzioni Animali, Facolta` di Agraria, Universita` degli studi di Catania, Catania, Italy 2 Laboratorio de Genetica, Facultad de Veterinaria, Universidad Complutense de Madrid, Madrid, Spain 3 DISTAFA, Facolta` di Agraria, Universita` degli studi Mediterranea, Polo Universitario di Feodivito, Reggio Calabria, Italy

Keywords Summary equine breeds; genetic diversity; microsatellites; molecular coancestry; The distribution of genetic diversity and the genetic relationships among population structure. western Mediterranean horse breeds were investigated using microsatellite markers. The examined sample included seven Spanish and three Correspondence Italian local horse breeds and populations, plus a Spanish Thoroughbred Prof. Donata Marletta, DACPA, sezione outgroup. The total number of animals examined was 682 (on average Scienze delle Produzioni Animali, Facolta` di 62 animals per breed; range 20–122). The microsatellite marker set ana- Agraria, Universita` degli studi di Catania, via Valdisavoia, 5 - 95123 Catania, Italy. Tel.: +39- lysed provided 128 alleles (10.7 alleles per locus). Within-breed genetic 95-234478; Fax: +39-95-234345; diversity was always high (>0.70), with breeds contributing about 8% of E-mail: [email protected] the total genetic variability. The mean molecular coancestry of the entire population examined was 0.205, Losino being the breed that con- Received: 5 April 2006; tributed most. In addition to Nei’s standard and Reynolds’ genetic dis- accepted: 10 June 2006 tances, pair-wise kinship distance and molecular coancestry were estimated. Remarkably similar breed rankings were obtained with all methods. Clustering analysis provided an accurate representation of the current genetic relationships among the breeds. Determining coancestry is useful for analysing genetic diversity distribution between and within breeds and provides a good framework for jointly analysing molecular markers and pedigree information. An integrated analysis was underta- ken to obtain information on the population dynamics in western Medi- terranean native horse populations, and to better determine conservation priorities.

In Spain, Aparicio (1960) reported two main Introduction equine types: the Celtic type of tarpanic origin in the In the Mediterranean area, the gradual replacement northern and western region of the peninsula, and of the horse by mechanization in agriculture has led the Spanish type, descendant from the African Barb to a severe reduction in the number of surviving horse, in the south and the east. The Spanish Celtic breeds. Of the 127 breeds in the Mediterranean horses, represented by a number of semi-feral local region, 23 no longer exist, 59 are endangered, 31 breeds and populations (Jaca Navarra, Asturcon, are not at risk and 14 are unknown (Nardone 2000). Caballo Gallego, Losina, Pottoka) with certain ‘prim- However, the Mediterranean region is home to a itive’ traits (such as below average size, different coat rich diversity of horse breeds. colours, etc.) are mainly bred in the north along the

ª 2006 The Authors Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 315 Genetic diversity in Mediterranean horses D. Marletta et al.

Cantabrian and show a clear genetic differentiation obtain coancestry coefficients that integrate the from the Mediterranean breeds, Menorquina and well-known genetic distances DS,DR and FST, tak- Mallorquina, native to the Balearic islands, which ing into account the allelic frequencies and diver- possess African and Arabian genes (Can˜ o´ n et al. sity of supposed founder populations. However, 2000). only a few authors have explored this approach for In Sicily, local ancient horse populations were analysing actual data sets (Caballero & Toro 2002; crossed with horses of diverse origin, including Ori- Fabuel et al. 2004). A valid alternative is a cluster- ental and African horses (Balbo 1995). Nowadays, ing method that uses individual multilocus geno- three local populations are bred in Sicily. The San- types to infer population structure and to assign fratellano breed, of which there are about 2000 individuals to theoretical populations (Pritchard head (including a selection nucleus of some 600 et al. 2000). animals), is reared in extensive systems in the The aim of this article was to assess the genetic mountainous north-east of the island. The Sicilian diversity in a set of 10 western Mediterranean horse Oriental Purebred, an endangered breed with less breeds and to determine their relationships by differ- than 70 animals remaining, represents the Sicilian ent methods. selection nucleus of the Arabian horse. A studbook has existed since 1923, but genealogical data are Materials and methods available from 1880. The Sicilian Indigenous is a very heterogeneous population. Largely unman- Sample collection aged, it probably originated from a primitive strain Blood samples were taken from seven Spanish and of Sicilian horse and from the Borbon Real Casa di three Italian local horse breeds and populations, and Ficuzza breed (related to the Neapolitan, Persano from a Spanish Thoroughbred outgroup. The entire and Arabian breeds). sample included 682 horses (122 Jaca Navarra, 119 These western Mediterranean populations show Asturcon, 72 Caballo Gallego, 66 Losina, 51 Pottoka, some common features: they are vigorous, well 31 Menorquina, 20 Mallorquina, 61 Sanfratellano, adapted to harsh conditions and stress-resistant. The 50 Sicilian Oriental Purebred, 30 Sicilian Indigenous size of most of their populations is small; some are and 60 Spanish Thoroughbred horses), all from threatened by extinction, and they have not been Spain and Sicily (Italy) (Table 1). systematically subjected to reproductive or breeding Animals for genotyping were chosen from large, technologies. As studbooks have only recently been well-defined geographical areas (with no considera- compiled, the possibility of genetic introgression tion of their possible genetic relationships), sampling between these populations cannot be ruled out. as many herds as possible in order to obtain a repre- According to FAO recommendations, determining sentative sample. In the case of the Sicilian Oriental classic genetic distances using neutral, highly poly- Purebred, the sample collected (n ¼ 50) consisted of morphic microsatellite markers is the method of nearly the entire studbook-registered population choice for investigating genetic relationships and (n ¼ 64). breed differentiation. This methodology also provides information for establishing preservation Microsatellite analysis priorities for livestock breeds (Barker 1999). Never- theless, phylogenetic methods based on genetic dis- To search for 12 microsatellite markers (HTG4, tances cannot be exhaustive in their quantification HTG6, HTG7, HTG10, HMS2, HMS3, HMS6, HMS7, of genetic diversity between closely related groups VHL20, ASB2, AHT4, AHT5), DNA was amplified in because they ignore within-group genetic variation three multiplex reactions according to standard pro- (Ruane 1999). Genetic diversity is a function of tocols using MJ Research PTC200 (MJ Research, relatedness between individuals that share one or Reno, NV, USA) and 9700 ABI Geneamp thermocy- more ancestors. The level of relatedness can be clers (Applied Biosystems, Foster City, CA, USA) (see expressed as a coefficient of kinship, which appears Can˜ o´ n et al. 2000). Fluorescent-labelled polymerase to be of central importance in the definition and chain reaction (PCR) products were diluted, mixed measurement of genetic diversity (Eding & Meuwis- with an internal size standard, and analysed using sen 2001). A molecular coancestry-based method ABI PRISM 3100 and ABI PRISM 377 genetic has recently been proposed for analysing the analysers, both equipped with Genescan and Ge- genetic diversity in subdivided populations (Cabal- notyper software (Applied Biosystems, Foster City, lero & Toro 2002). This uses molecular data to CA, USA). Seven reference DNA samples were used

ª 2006 The Authors 316 Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 D. Marletta et al. Genetic diversity in Mediterranean horses

Table 1 Name, origin, population size, studbook existence, number (n) of samples, expected heterozygosities (He), FIS values, mean number of alleles (MNA), allelic richness (AR) for each of the eleven horse breeds across all 12 microsatellite loci

Breed name Country Population size Existence of Studbook (year) Sample size He FIS MNA AR

Asturcon Spain 1000 Yes (1981) 119 0.733 )0.009 7.17 5.8 Jaca Navarra Spain 250 No 122 0.729 0.035* 7.83 6.2 Losino Spain <200 No 66 0.704 )0.022 6.83 5.9 Caballo Gallego Spain >10 000 No 72 0.761 0.060* 7.92 7.0 Pottoka Spain 1000 Yes (1995) 51 0.775 0.042* 7.08 6.6 Menorquin Spain 1400 Yes (1993) 31 0.721 )0.040 6.25 6.0 Mallorquin Spain <200 Yes (1993) 20 0.748 0.054* 6.08 6.1 Sanfratellano Italy <700 No 61 0.751 0.019 6.92 6.2 Sicilian Oriental Italy <70 Yes (1923) 50 0.702 )0.003 6.67 5.7 Sicilian Indigenous Italy 400 No 30 0.803 0.098* 7.92 7.4 Spanish Thoroughbred Spain 60 0.690 )0.029 5.17 4.7

*Values different from 0 at p < 0.05.

to ensure compatibility between the results obtained between and within breeds were calculated based on at the two participating laboratories. the molecular coancestry concept (Caballero & Toro 2002). The molecular coancestry (f) between two individuals (say i and j) at a given locus is expressed Statistical analysis as 1/4[I11 + I12 + I21 + I22], where Ixy ¼ 1 when Unbiased estimates of gene diversity (expected allele x in the first individual and allele y on the heterozygosity or Hardy–Weinberg heterozygosity), same locus in the second individual are identical, observed heterozygosity, and the number of alleles and 0 when they are not. For L number of loci, per breed (the last two with their associated standard molecular coancestry is obtained by simply dividing errors), were calculated using the Microsatellite by the number of analysed loci P Toolkit (Park 2001). The FSTAT program (Goudet L 2001) was used for testing the conformity of geno- i¼1 fij;l fij ¼ : type frequencies to the Hardy–Weinberg equilib- L rium. Sequential Bonferroni analysis (Rice 1989) The molecular coancestry of an individual i with was used to achieve a global type I error of 0.05. itself is the self-coancestry (si), which is related to FSTAT software was also used to calculate the allelic homozygosity (Fi) by the expression Fi ¼ 2si ) 1. richness (AR) standardized for variation in sample Genetic diversity can be partitioned as: size. When calculated in this way, the number of al- ~ ~ leles between populations with different sample sizes ð1 f Þ¼ð1 ~sÞþð~s f Þþðf f Þ can be compared. where Wright’s F coefficients (F , F and F ) were cal- P P IT IS ST n n fij culated using Genetix 4.0 software (Belkhir et al. i¼1 j¼1 ; f ðthe average global coancestryÞ¼ 2 2001) The significance of the FIT and FIS values were n tested by permuting alleles 1000 times within the set Pn of breeds or within each breed, respectively. Signifi- fii cant deviation of FST from the null hypothesis was ~f ¼ i¼1 tested for by random permutations of genotypes n among samples. and Two classical measures of genetic distance were P n s made using Population software (Langella 2002): the ~s ¼ n¼1 i ; n standard Nei (DS) distance, which increases linearly with time, and the Reynolds (DR) distance, which and n the number of breeds. under a pure genetic drift model (excluding muta- The average kinship distance for breed i is: Dii ¼ tions and admixtures) also increases linearly with (si ) fii). The proportion of diversity between individ- time. In addition, genetic diversity components uals for breed i is: Gi ¼ Dii/(1 ) fii). The total genetic

ª 2006 The Authors Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 317 Genetic diversity in Mediterranean horses D. Marletta et al.

diversity GDT ¼ð1 f Þ is partitioned into three etic composition, which in practice is only likely to components: the genetic diversity within individuals happen if i and j are the same population. Distances [GDWI ¼ð1 ~sÞ], the genetic diversity between were averaged over five runs at K ¼ 10 and repre- ~ individuals [GDBI ¼ð~s f Þ], and the genetic diver- sented on a splits-graph based on the split-decompo- ~ sity between breeds [GDBS ¼ðf f Þ]. sition method (Dress et al. 1996); this routine is part Finally, the kinship between two individuals i and of the SplitsTree program (Huson & Bryant 2006). j was used as a measure of genetic distance among individuals within breeds or among breeds, calcula- Results ted as Dij ¼ ([si + sj]/2) ) fij (Caballero & Toro 2002), and the proportional contribution of each breed to Genetic diversity within breeds the estimated global coancestry. Table 2 shows the number of alleles for each micro- The number of genetic clusters maximizing the satellite marker, which ranged from 6 to 14 likelihood of the molecular data, and the proportion (average 10.7). The observed and the expected of the individual’s genome derived from each heterozygosities for each marker are also given. inferred theoretical ancestral population or cluster, Table 1 shows the average number of alleles and were estimated using Structure software (Pritchard the AR of each breed. The Sicilian Indigenous pop- et al. 2000). This model-based clustering program ulation showed the highest values for all these vari- uses a Monte Carlo Markov chain (MCMC) algo- ables; the Spanish Thoroughbred showed the rithm to assess the presence of a structure underly- lowest. Table 1 also shows the values for the expec- ing the genetic information provided by the genetic ted heterozygosity under Hardy–Weinberg equilib- markers. A genetic distance between the breeds was rium and the measure of the possible discrepancy defined using the genetic composition output from between this value and the observed heterozygosity this program. If qk(i) is the proportion of the genome (FIS). The negative FIS values of some breeds indi- of breed i that comes from population k, then the cate an excess of heterozygous genotypes with distance between breeds i and j can be defined as respect to the expected value. The FIS values for (Can˜ o´ n et al. 2006): the rest of the breeds, even if statistically signifi- XK cant, are not far from 0, indicating that mating is qkðiÞþqkðjÞ dsði; jÞ :¼ jqkðiÞqkðjÞj : close to panmixia. 2 k¼1 Table 3 shows the values for the molecular self-co-

Thus, two breeds are genetically closer when the ancestry of each breed (fii), the average individual proportion of their shared genomes is higher and self-coancestries (si), the within-breed kinship distan- when these have an important weighting in the ﬁnal ces (Dii), the average inbreeding (Fi), and the pro- genetic composition. It is obvious that dS(i,j) ‡ 0, portion of diversity between individuals(Gi). The that dS(i,j) ¼ dS(j,i), dS(i,i) ¼ 0, and that dS(i,j) ¼ 0 table highlights two main points: the similarity of implies that i and j have exactly the same mean gen- the values for many of these parameters, as well as

Table 2 Number of alleles, range of allele sizes, F-statistics (FST, and FIS), and expected (He) and observed (Ho) heterozygosities for each of the 12 microsatellite markers in 11 European horse breeds

Locus No. of alleles Range allele sizes FST FIS Ho He

HTG-4 10 116–138 0.078 )0.011 0.698 0.692 HTG-6 12 82–106 0.138 0.010 0.630 0.634 HTG-7 6 117–127 0.094 )0.018 0.614 0.608 HTG-10 12 86–108 0.093 0.061 0.720 0.798 HMS-2 13 216–240 0.131 0.017 0.749 0.750 HMS-3 10 148–170 0.136 0.142 0.634 0.728 HMS-6 9 152–174 0.135 0.033 0.661 0.689 HMS-7 9 166–182 0.066 )0.036 0.810 0.784 VHL20 13 81–111 0.057 )0.008 0.809 0.798 ASB-2 14 216–250 0.057 0.005 0.777 0.799 AHT-4 11 137–159 0.075 )0.010 0.804 0.808 AHT-5 9 124–140 0.064 0.014 0.768 0.767

ª 2006 The Authors 318 Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 D. Marletta et al. Genetic diversity in Mediterranean horses

Table 3 Molecular self-coancestry (fii), average of individual self-coancestry (si), within-breed kinship distance Dii, inbreeding (Fi), proportion of diversity between individuals (Gi) and proportional contribution of each breed i to the global coancestry of the 10 Mediterranean horse breeds included in the study

Distance Distance to

fii si Dii Fi Gi within breed other breeds Total

Asturcon 0.271 0.631 0.360 0.261 0.493 0.0271 0.0070 0.0201 Jaca Navarra 0.275 0.648 0.374 0.297 0.515 0.0275 0.0065 0.0210 Losino 0.302 0.640 0.339 0.281 0.485 0.0302 0.0071 0.0230 Caballo Gallego 0.244 0.642 0.398 0.284 0.527 0.0244 0.0050 0.0194 Pottoka 0.233 0.629 0.396 0.258 0.516 0.0233 0.0051 0.0182 Menorquin 0.290 0.625 0.335 0.250 0.472 0.0290 0.0075 0.0216 Mallorquin 0.270 0.646 0.376 0.292 0.515 0.0270 0.0089 0.0182 Sanfratellano 0.255 0.631 0.376 0.262 0.504 0.0255 0.0069 0.0186 Sicilian Oriental 0.305 0.648 0.343 0.296 0.493 0.0305 0.0114 0.0191 Sicilian Indigenous 0.210 0.638 0.427 0.276 0.541 0.0210 0.0065 0.0145

the coincidence in the fii and Fi values in most diversity (0.795) can be distributed into three com- breeds. These features, together with Gi values close ponents: the genetic diversity between breeds to 0.5 for many of the breeds, indicate random mat- (GDBS ¼ 0.063), the genetic diversity within individ- ing practices. In the Asturcon, Losino, Menorquina uals (GDWI ¼ 0.362), and the genetic diversity and Oriental Purebred breeds, the Gi values were between individuals (GDBI ¼ 0.370). The sum of the lower than 0.5, showing variability to be preferen- last two components represents the genetic diversity tially distributed within individuals. within breeds (0.732). Expressing these values as Table 3 also shows the proportional contribution percentages provides the corresponding F values: of each breed to the global coancestry of these FIS ¼ 0.012, FST ¼ 0.079 and FIT ¼ 0.091. horses. The Sicilian Oriental Purebred and Losino breeds were found to contribute the most to global Genetic distances and clustering coancestry, a consequence of their own coancestry values. However, the total contribution of Losino is Together, Tables 4 and 5 show the Nei’s standard higher because of its closer relationships with the distances, the Reynolds’ genetic distances and the other breeds. pairwise kinship distance and molecular coancestry Excluding the Spanish Thoroughbred, the values matrix. The genetic distances for the Spanish Thor- of 0.277 for mean coancestry, 0.268 for average oughbred were greater than for any other breed. inbreeding, 0.638 for average self-coancestry, and The Oriental Purebred and Mallorquina breeds 0.205 for mean coancestry in the whole population, were, on average, the most distant breeds, inde- summarize the genetic diversity. The total genetic pendent of the genetic distance measurement used.

Table 4 Nei’s standard DS (above the diagonal) and Reynolds’ DR genetic distances (below the diagonal) among 10 Mediterranean horse breeds

Caballo Sicilian Sicilian Asturcon Jaca Navarra Losino Gallego Pottoka Menorquin Mallorquin Sanfratellano Oriental Indigenous

Asturcon 0.224 0.221 0.151 0.227 0.394 0.449 0.368 0.712 0.392 Jaca Navarra 0.068 0.192 0.127 0.164 0.275 0.462 0.358 0.765 0.354 Losino 0.071 0.063 0.160 0.232 0.365 0.520 0.342 0.622 0.360 Caballo Gallego 0.042 0.036 0.048 0.125 0.188 0.312 0.268 0.595 0.319 Pottoka 0.061 0.045 0.067 0.028 0.151 0.243 0.301 0.615 0.281 Menorquin 0.109 0.080 0.110 0.049 0.037 0.278 0.483 0.649 0.484 Mallorquin 0.112 0.116 0.137 0.072 0.053 0.070 0.596 0.790 0.560 Sanfratellano 0.099 0.097 0.099 0.068 0.072 0.122 0.132 0.342 0.152 Sicilian Oriental 0.179 0.189 0.175 0.148 0.148 0.172 0.185 0.098 0.278 Sicilian Indigenous 0.091 0.085 0.093 0.067 0.055 0.107 0.107 0.031 0.073

ª 2006 The Authors Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 319 Genetic diversity in Mediterranean horses D. Marletta et al.

Table 5 Pairwise kinship distance (above diagonal) and molecular coancestry (below diagonal) among 10 Mediterranean horse breeds

Jaca Caballo Sicilian Sicilian Asturcon Navarra Losino Gallego Pottoka Menorquin Mallorquin Sanfratellano Oriental Indigenous

Asturcon 0.421 0.406 0.416 0.430 0.439 0.466 0.448 0.498 0.473 Jaca Navarra 0.218 0.407 0.417 0.424 0.422 0.475 0.453 0.513 0.474 Losino 0.229 0.237 0.410 0.425 0.427 0.473 0.438 0.481 0.463 Caballo Gallego 0.221 0.228 0.231 0.425 0.413 0.456 0.444 0.494 0.475 Pottoka 0.200 0.215 0.210 0.210 0.403 0.441 0.448 0.495 0.466 Menorquin 0.189 0.215 0.205 0.220 0.224 0.423 0.457 0.481 0.479 Mallorquin 0.172 0.172 0.170 0.188 0.197 0.212 0.492 0.516 0.506 Sanfratellano 0.183 0.186 0.198 0.193 0.182 0.171 0.146 0.440 0.434 Sicilian Oriental 0.141 0.135 0.163 0.151 0.144 0.155 0.130 0.200 0.451 Sicilian Indigenous 0.161 0.169 0.176 0.165 0.167 0.152 0.136 0.200 0.192

The Caballo Gallego breed was the most similar to breeds, and a very low kinship distance between the all the others when classical genetic distances were Pottoka and Menorquino breeds. These relationships, calculated, and the Losino breed the most similar shown in the dendrogram in Figure 1, show the when coancestry-based genetic distances were breeds to group into two main clusters: the Italian used. The similarity of breed ranking obtained with and Spanish groups. The length of the branches cor- all the genetic distance methods is remarkable responds to the level of inbreeding; the Sicilian Ori- (Table 6). ental Purebred, Mallorquino and Menorquino breeds The picture provided by kinship distances, calcula- therefore appear to have suffered higher levels of ted by using the molecular coancestry concept, inbreeding than the rest. shows a high level of coancestry and a low genetic This value is the result of assuming no migration distance (Dk) between the Losino and Jaca Navarra between breeds after their reproductive isolation. The information obtained with the Bayesian model- based procedure, however, gives an idea of gene ﬂow between breeds. In fact, a way to see this differ- Table 6 Product–moment correlation for distance matrices compar- entiation (or any kind of phylogeny among the ison: molecular coancestry (f), Nei’s standard (DS), Reynolds’ (DR), and breeds analysed) is to examine the population struc- kinship distance (Dk) ture using the clustering model-based method (Pritchard et al. 2000), and assuming different K val- Distance DS fDk ues (i.e. the number of clusters). Using this method, f )0.891 the Italian and Spanish breeds always form clusters ) Dk 0.902 0.986 of their own, indicating a high degree of isolation D 0.969 )0.773 0.796 R (Figure 2). As K increases, new subgroups arise, but

Figure 1 Neighbour-joining tree based on kinship distances for the 10 Mediterranean horse breeds (numbers refer to the fraction of bootstraps, greater than 50%, supporting each branch).

ª 2006 The Authors 320 Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 D. Marletta et al. Genetic diversity in Mediterranean horses

K = 2

K = 3

K = 4

100%

50 K = 5 25

Figure 2 Percentage membership of each

breed to each of the possible clusters, assu- Losino Pottoka ming k ¼ 2 to 5. Each colour or grid repre- Asturcon Menorquin Mallorquin Sanfratellano sents the percentage membership of each Jaca_Navarra Caballo_Gallego breed, represented by a vertical parallelo- Sicilian_Oriental

gram. Sicilian_Indigenous

Table 7 Proportion of memberships of each breed assigned to each one of the 10 possible clusters

Inferred clusters

Breed 12345678910

Asturcon 0.453 0.019 0.344 0.042 0.014 0.016 0.030 0.017 0.007 0.057 Jaca Navarra 0.027 0.252 0.028 0.260 0.028 0.046 0.283 0.030 0.008 0.039 Losino 0.025 0.027 0.028 0.019 0.032 0.029 0.033 0.752 0.012 0.042 Caballo Gallego 0.040 0.038 0.089 0.059 0.069 0.098 0.124 0.079 0.016 0.388 Pottoka 0.028 0.033 0.031 0.062 0.288 0.221 0.073 0.062 0.017 0.186 Menorquin 0.014 0.034 0.019 0.019 0.741 0.078 0.032 0.025 0.008 0.031 Mallorquin 0.012 0.009 0.030 0.007 0.275 0.561 0.016 0.013 0.011 0.066 Sanfratellano 0.022 0.051 0.030 0.017 0.013 0.047 0.040 0.028 0.711 0.042 Sicilian Oriental 0.010 0.007 0.008 0.007 0.017 0.021 0.011 0.016 0.869 0.037 Sicilian Indigenous 0.035 0.064 0.018 0.035 0.018 0.100 0.046 0.049 0.577 0.058

ª 2006 The Authors Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 321 Genetic diversity in Mediterranean horses D. Marletta et al.

Figure 3 Splits-graph for the 10 Mediterra- nean horse populations based upon the

split-decomposition method and on qk(i) (see Materials and methods).

only because the genomes of the Spanish breeds observed in these 10 native populations (7.07) was become distributed across the different clusters. the same as that observed for Lipizzan horse from When the number of clusters assumed is the same seven national studs (Achmann et al. 2004), greater as the number of breeds (10), the results shown in than that observed in 11 Asian and two European Table 7 are obtained. It can be seen that the Italian populations (Tozaki et al. 2001), in Thoroughbred breed genomes group fall within the same cluster, from four national studs, but one (Cunningham et al. while those of the Spanish breeds, with the excep- 2001) and in other European breeds (Aberle et al. tion of those of the Losino and Menorquino breeds 2004; Glowatzki-Mullis et al. 2005).The genetic [which are mainly assigned to their own cluster diversity was high (>0.70), except for the Spanish (75.2% and 74.1%, respectively)], belong to two Thoroughbred, and comparable with that of other or three different clusters. Even so, the horses of horse breeds (Wimmers et al. 1998; Bjørnstad & the Balearic Islands share a main ancestral popula- Røed 2001; Cunningham et al. 2001). tion with the Pottoka breed (Table 7). Figure 3 As expected, the more intensive and effective the shows the relationships among all breeds based on selection, the less genetic variability. The Spanish the proportion of the genome shared across clusters Thoroughbred showed the lowest gene diversity val- using a splits-graph. The goodness-of-ﬁt estimate ues (0.690), followed by the Oriental Purebred for this splits-graph (a measure of the accuracy of (0.702) among the Sicilian populations, and the the representation of the data set) is particularly Losina breed (0.704) among the Spanish ones. Both good (97.5%). In this type of graph, the sum of these breeds have very small population sizes. the lengths of all the edges along the shortest Heterozygosity in these was higher than that seen in path from one breed to another is proportional to other endangered European and Asiatic breeds the actual distance between those breeds in the (Iriondo et al. 2002; Tozaki et al. 2003). data set. It is noteworthy that, despite the small sample size (n ¼ 30), the Sicilian Indigenous population showed the greatest genetic variability, probably because of Discussion its wide genetic base and heterogeneity (which is The total genetic diversity of the analysed horse pop- apparent in its morphology). For all the breeds, gen- ulations was high (0.795) and mostly (92%) etic diversity was consistent with allele richness retained within breeds. The number of alleles rank, i.e. it was independent of the sample size.

ª 2006 The Authors 322 Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 D. Marletta et al. Genetic diversity in Mediterranean horses

The coefficient FIS, which indicates the degree of cestry when trying to explain the genetic distances departure from random mating, showed a small defi- between breeds. cit of heterozygotes (1.63%) as revealed by a single The strong coancestry and low Dk between the marker (HMS3). At the population level, the negat- Losino and Jaca Navarra breeds suggest recent dif- ive FIS value could be explained by the existence of ferentiation. The reduced kinship distance between subpopulations which were pooled together to form the Pottoka and Menorquina breeds indicate that the present populations (Wahlund effect), and/or by they have received genetically similar migrants an asymmetrical sex migration that produced an rather than having followed a common evolution- outbreeding effect in the progeny. This would have ary path. resulted in an excess of heterozygotes. This is quite The large genetic distance between the Sicilian noticeable in the Menorquina, Spanish Thorough- and Balearic breeds was unexpected. Both breeds bred and Losina breeds, but less clear in the Astur- were influenced by the Arabian horse (Can˜ o´ n et al. con and Oriental Purebred breeds. In these 2000) and belong to the Mediterranean trunk. populations, Fi is lower than fii, and Gi is lower than Nonetheless, the Oriental Purebred and Mallorquino 0.5, showing that genetic variability is preferentially breeds appear to be the most divergent (in terms of distributed within individuals. The negative values their Ds, DR and Dk values) and to show the lowest could also be explained as the effect of a breeding coancestry value (fii). This could be due to the strategy that avoids mating between closely related absence of migration, or reproductive isolation and animals. consequent genetic drift. Nowadays, the Mallorquino

Positive FIS values mean a significant deficit of and Menorquino breeds are probably the products of heterozygotes. Such values were shown by the Sicil- the genetic drift caused by their own geographical ian Indigenous, Caballo Gallego, Pottoka, Mallorqu- and reproductive isolation. Can˜ o´ n et al. (2000) ina, and Jaca Navarra breeds. This, plus the related reported that the Mallorquino, Menorquino and discrepancy between fii and Fi, might be explained Losino breeds show a clear clustering among the by subdivisions in these breeds, and might even be Iberian horses; this is confirmed in the present work partially due to samples from genetically different by their high level of molecular coancestry and by subpopulations being obtained. Furthermore, the results of the Structure software analysis. because of the breeding systems followed, the estab- When assuming the same number of clusters and lishment of male lineages in each breed cannot be breeds, cluster analysis showed that, on average, ruled out. most of the genomes of the breeds with small popu- The coancestry values seemed to be quite similar lation sizes (characterized by the highest coancestry for all the tested breeds. The most important data values, i.e. the Sicilian, Mallorquino, Menorquino were the high value of fii and the degree of inbreed- and Losino breeds), grouped into a single cluster. ing (Fi) in the endangered Oriental Purebred breed. The Balearic breeds appear to partially share a clus- In addition, the Sicilian Indigenous breed showed ter, whereas most of the Sicilian horses belong to high Dii and Gi values, and therefore the greatest the same cluster (meaning they derive from a single genetic variation. This is probably a reflection of a ancestral population). Moreover, it should be noted still heterogeneous population. that the splits-graph (Figure 3) shows a quite tree-

The pairwise genetic distances estimated by DS and like form (Figure 1), revealing different evolutionary DR, the methods of choice for assessing relationships pressures and/or the origin of migrants to the Balea- among breeds, were virtually the same. The classic ric and Sicilian breeds. This is probably because of genetic distances DS and DR are a consequence of the common history of Sicilian horses, all of which the change in the degree of inbreeding after breed were influenced by Oriental breed genes. Despite the separation. Coancestry (f), however, is a measure of short pedigrees recorded, Oriental Stallions were similarity, and the kinship distance (DK) takes into admired for their properties and were bred in Sicily account the allele frequencies in the founder popula- for many centuries. This surely influenced the Sicil- tion (Eding & Meuwissen 2001; Eding et al. 2002). ian Oriental Purebred and other native populations. The correlations between all the genetic distances The present paper only considers information pro- measured for the studied horse breeds were highly vided by neutral markers. These are of interest for significant. This might be explained by a smaller estimating population histories and for inferring rela- influence of the allele frequencies in the founder tionships, but in conservation campaigns, additional population on kinship distances. This finding high- genetic variability based on specific loci related with lights the importance of estimating molecular coan- disease resistance or adaptive features, as well as

ª 2006 The Authors Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 323 Genetic diversity in Mediterranean horses D. Marletta et al. information on traits of economic impact must be Belkhir K., Borsa P., Chikhi L., Raufaste N., Bonhomme considered jointly. The use of molecular coancestry F. (2001) Genetix, logiciel sous Windows TM pour la geńe´- for measuring genetic diversity is appealing because tique des populations, Laboratoire Geńome, Populations, the interrelation between these parameters and oth- Interactions, CNRS UPR 9060, Universite´ de Montpellier II, ers based on pedigree information is immediately Montpellier (France). http://www.genetix.univ-montp2. provided. Thus, the integration of both types of fr/genetix/genetix.htm information in the same framework is feasible (see Bjørnstad G., Røed K.H. (2001) Breed demarcation Caballero & Toro 2002). Genetic markers are a use- and potential for breed allocation of horses assessed ful tool for estimating coancestry in the absence of by microsatellite markers. Anim. Genet., 32, 59– 65. pedigree information – the usual situation when Caballero A., Toro M.A., (2002) Analysis of genetic considering unconnected populations or breeds. diversity for the management of conserved subdivided However, within a breed it is also usual to have populations. Conserv. Genet., 3, 289–299. pedigree data, which might better help estimate gen- Can˜ o´ n J., Checa M.l., Carleos C., Vega-Pla J.l., Vallejo ealogical coancestry. M., Dunner S., (2000) The genetic structure of Spanish In the light of the available data for western Medi- Celtic horse breeds inferred from microsatellite data. terranean native horse populations, classical genetic Anim. Genet., 31, 39–48. distances tend to overestimate the genetic diversity Can˜ o´ n J., Garcıá D., Garcıá-Atance M.A., Obexer-Ruff in inbred populations, in which genetic drift causes G., Lenstra J.A., Ajmone-Marsan P., Dunner S., the extremely different allele frequencies. This contrasts ECONOGENE Consortium (2006) Geographical parti- with that seen in large, non-inbred, populations tioning of goat diversity in Europe and the Middle characterized by their even distributions of allele fre- East. Anim. Genet., 37, 327–334. quencies. Therefore, genetic diversity information, Cunningham E.P., Doley J.J., Splan R.K., Bradley D.G. evaluated by integrating within- and between-popu- (2001) Microsatellite diversity, pedigree related- lation analyses may allow conservation priorities to ness and the contributions of the founder lineages be better established. to thoroughbred horses. Anim. Gen., 32, 360–364. Dress A.W.M., Huson D.H., Moulton V. (1996) Analyzing and visualizing sequence and distance data using Acknowledgements SplitsTree. Discrete Appl. Math., 71, 95–109. We are grateful to Jose´ L. Vega-Pla who provided Eding H., Meuwissen T.H.E. (2001) Marker-based esti- samples from some of the Spanish breeds. Financial mates of between and within population kinships for support was provided by ACPRA, a grant AGF95- the conservation of genetic diversity. J. Anim. Breed. 064 and by FONDO 6/F Assistenza Tecnica in Genet., 118, 141–159. Eding H., Crooijmans R.P.M.A., Groenen M.A.M., Meu- Zootecnia. wissen T.H.E. (2002) Assessing the contribution of breeds to genetic diversity in conservation schemes. References Genet. Sel. Evol., 34, 613–633. Fabuel E., Barragań C., Silio´ , L., Rodrı´guez M.C., Toro Aberle K.S., Haman H., Dro¨ gemu¨ ller C., Distl O. (2004) M.A. (2004) Analysis of genetic diversity and conserva- Genetic diversity in German draught horse breeds com- tion priorities in Iberian pigs based on microsatellite pared with a group of primitive, riding and wild horses markers. Heredity, 93, 104–113. by means of microsatellite DNA markers. Anim. Genet., Glowatzki-Mullis M.L., Muntwyler J., Pfister W., Marti E., 35, 270–277. Rieder S., Poncet P.A., Gaillard C. (2005) Genetic diver- Achmann R., Curik I., Dovc P., Kavar T., Bodo I., Habe sity among horse populations with a special focus on F., Marti E., Solkner J., Brem G. (2004) Microsatellite the Franches-Montagnes breed. Anim. Gen., 37, 33–39. diversity, population subdivision and gene flow in the Goudet J. (2001) FSTAT, a program to estimate and test Lipizzan horse. Anim. Gen., 35, 285–292. gene diversities and fixation indices (version 2.9.3). Aparicio G. (1960) Zootecnia Especial. Imprenta Moderna. http://www2.unil.ch/popgen/softwares/fstat.htm Co´ rdoba, Spain. Huson D.H., Bryant D. (2006) Application of phylogenet- Balbo S. M. (1995) L’influenza dell’arabo-orientale sul ic networks in evolutionary studies. Mol. Biol. Evol., 23, cavallo siciliano. In: Savier M. (ed.), L’Asil Arabo - Il 254–267. cavallo nobile d’Arabia. R&R Editrice, Spoleto (PG), Iriondo M., Jugo B., Rendo F., Solis A., Mazon L.I., Italy, pp. 169–175. Estonba A. (2002) Analysis of genetic diversity of indi- Barker J.S.F. (1999) Conservation of livestock breed genous cattle, horse and sheep breeds from the Basque diversity. Anim. Genet. Res. Inf., 25, 33–43.

ª 2006 The Authors 324 Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 D. Marletta et al. Genetic diversity in Mediterranean horses

country. In:Proceedings of the 7th World Congress on Rice W.R. (1989) Analyzing tables of statistical tests. Evo- Genetics Applied to Livestock Production. Montpellier, lution, 43, 223–225. France, 19–23 August, 2002. Ruane J., (1999) A critical review of the values of genetic Langella O. (2002) Population 1.2.28. Logiciel de distance studies in conservation of animal genetic geńe´tique des populations. Laborotoire Populations, resources. J. Anim. Breed. Genet., 116, 767–779. geńe´tique et e´volution, CNRS UPR 9034, Gif-sur- Tozaki T., Kakoi H., Mashima S., Hirota K., Hasegawa T., Yvette. http://www.legs.cnrs-gif.fr Ishida N., Miura N., Choi-Miura N.H., Tomita M. Nardone A. (2000) Weather conditions and genetics of (2001) Population study and validation of paternity breeding systems in the Mediterranean area. In:Pro- testing for Thoroughbred horses by 15 microsatellite ceedings XXXV Simposio Internazionale di Zootecnica. loci. J. Vet. Med. Sci., 63, 1191–1197. Produzioni animali di qualita` ed impatto ambientale Tozaki T., Takezaki N., Hasegawa T., Ishida N., Kurosawa nel sistema mediterraneo, Ragusa, Italy, 25 May 2000. M., Tomita M., Saitou N., Mukoyama H. (2003) Micro- pp. 67–91. satellite variation in Japanese and Asian horses and Nei M., (1987) Molecular Evolutionary Genetics. Columbia their phylogenetic relationship using a European horse University Press, New York. outgroup. J. Hered., 94, 374–380. Park S.D.E. (2001) Trypanotolerance in West African Cattle Wimmers K., Ponksuksil S., Schmoll F. Hardge T., Hatzi- and the Population Genetic Effects of Selection, PhD thesis, panagiotu A., Weber J., Wostmann S., Olek K., Schel- University of Dublin, Dublin, Ireland. lander K. (1998) Efficiency of microsatellite markers of Pritchard J.K., Stephens M., Donnelly P. (2000) Inference the international standard panel for parentage control of population structure using multilocus genotype data. in German horse populations. Zuchtungskunde, 70, 233– Genetics, 55, 945–959. 241.

ª 2006 The Authors Journal compilation ª 2006 Blackwell Verlag, Berlin • J. Anim. Breed. Genet. 123 (2006) 315–325 325